Three dimensional motion compensated video coding

• US 4,999,705 A
• Filed: 05/03/1990
• Issued: 03/12/1991
• Est. Priority Date: 05/03/1990
• Status: Expired due to Term

• Current Assignee: American Telephone & Telegraph Company (AT&T, Inc.)
• Original Assignee: AT&T, Inc.
• Inventors: Puri, Atul
• Primary Examiner(s): Kostak, Victor R.
• Application Number:
  US07/518,677
• Time in Patent Office: 313 Days
• Field of Search: 358/133, 358/135, 358/136, 358/105, 382/56
• US Class Current: 348/412.1
• CPC Class Codes: H04N 19/00 Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
• CPC Class Codes: H04N 19/51 Motion estimation or motion compensation
• CPC Class Codes: H04N 19/577 Motion compensation with bidirectional frame interpolation, i.e. using B-pictures
• CPC Class Codes: H04N 19/60 using transform coding
• CPC Class Codes: H04N 19/61 in combination with predictive coding
• CPC Class Codes: H04N 19/62 by frequency transforming in three dimensions H04N19/63 takes precedence
• CPC Class Codes: H04N 19/649 the transform being applied to non rectangular image segments

Abstract

A video signal encoder uses three dimensional transform coding on blocks of intensity values and then selects the most significant coefficients for further processing by separating the coefficients into several groups have approximately the same energy and thus approximately the same significance. The intensity values can be error values computed by forming the difference between the actual intensity values and values computed using conditional motion compensated interpolation or motion compensated prediction. The encoder can use both two dimensional and three dimensional transformations, and if desired, adaptively switch between those processing techniques. A complementary decoder recovers the intensity values from encoded signals.

Claims

1. A method of encoding a signal representing the actual intensity values of picture elements (pels) in a series of frames, comprising the steps of

(i) computing estimates of the values of pels in at least two intermediate frames which lie between reference frames, based on the values of pels in at least one of said reference frames;

(ii) combining error values for corresponding blocks of pels in each of said intermediate frames to form a series of transform coefficients representing said error values, each of said error values representing the difference between said estimated value and said actual value of said pel, and

(iii) selecting ones of said coefficients for further processing based upon the significance of information represented by said coefficients,

wherein said selection step includes

(a) grouping said coefficients into a plurality of groups each containing coefficients having approximately the same significance, and

(b) selecting all coefficients from only selected ones of said groups.

2. The invention defined in claim 1, wherein said computing step includes

(a) computing estimates of motion of objects in said frames, and

(b) predicting the values of said pels based upon said motion estimates and the values of corresponding pels in at least one of said reference frames.

3. The invention defined in claim 1, wherein said computing step includes

(a) computing estimates of motion of objects in said frames, and

(b) interpolating the values of said pels based upon said motion estimates and the values of corresponding pels in at least two of said reference frames.

4. The invention defined in claim 1

wherein said computing step includes computing estimates for P intermediate frames, each of said P frames including an M×

N block of pels, and

wherein said grouping step includes forming up to S=M+N+P-2 groups of coefficients.

8. The invention defined in claims 1 or 5, wherein said combining step includes forming discrete cosine transform coefficient c(j,k,l) for each of said blocks of pels s(m,n,p).

9. The invention defined in claim 8 wherein said block of pels is of size M×

N×

P, and wherein;

##EQU8##

5. A method of encoding a video signal representing the actual intensity values of picture elements (pels) in a series of frames, comprising the steps of

(i) computing estimates of the intensity values of pels in at least two intermediate frames which lie between reference frames, based on the intensity values of pels in at least one of said reference frames;

(ii) combining error values for corresponding blocks of pels in each of said intermediate frames to form a series of transform coefficients representing said error values, each of said error values representing the difference between said estimated intensity value and said actual intensity value of said pel,

(iii) selecting ones of said coefficients for further processing based upon selected characteristics of said coefficients, and

(iv) encoding said selected coefficients to produce a representation of said video signal,

wherein said intensity estimiate computing step includes

(a) forming estimates of the motion of objects in said frames, and

(b) calculating said intensity estimates as a function of said motion estimates.

6. The invention defined in claim 5, wherein said intensity estimates calculation is formed using motion compensated prediction.

7. The invention defined in claim 5, wherein said intensity estimates calculation is formed using conditional motion compensated interpolation.

10. A method of encoding a signal representing the intensity values of picture elements (pels) in a series of frames, comprising the steps of computing error values s(m,n,p) for three dimensional blocks of pels, said 3D blocks including M×

N pels in each of P of said frames;

forming M×

N×

P discrete cosine transformation coefficients c(j,k,l) for each of said 3D blocks of error values;

grouping said coefficients into S groups, each of said groups containing coefficients having approximately the same energy, and

selecting coefficients only from ones of said groups to represent said signal.

11. The method of claim 10, wherein said computing step includes forming said error values using conditional motion compensated interpolation (CMCI).

12. The method of claim 10, wherein said computing step includes forming said error values using motion compensated prediction (MCP).

13. The invention defined in claims 11 or 12 wherein said transformation coefficients c(j,k,l) are given by:

##EQU9##

14. The invention defined in claim 13 wherein M=N=8, P=4, and wherein S=18 groups C₀ through C₁₇, said groups having coefficients as follows:

space="preserve" listing-type="tabular">______________________________________ Group Number of Coefficients ______________________________________ C.sub.0 1 C.sub.1 3 C.sub.2 6 C.sub.3 10 C.sub.4 14 C.sub.5 18 C.sub.6 22 C.sub.7 26 C.sub.8 28 C.sub.9 28 .sup. C.sub.10 26 .sup. C.sub.11 22 .sup. C.sub.12 18 .sup. C.sub.13 14 .sup. C.sub.14 10 .sup. C.sub.15 6 .sup. C.sub.16 3 .sup. C.sub.17 1 ______________________________________

15. Apparatus for encoding a signal representing the actual intensity values of picture elements (pels) in a series of frames, comprising

(i) means for computing estimates of the values of pels in at least two intermediate frames which lie between reference frames, based on the values of pels in at least one of said reference frames;

(ii) means for combining error values for corresponding blocks of pels in each of said intermediate frames to form a series of transform coefficients representing said error values, each of said error values representing the difference between said estimated value and said actual value of said pel, and

(iii) means for selecting one of said coefficients for further processing based upon the significance of information represented by said coefficients,

wherein said selection means includes means for grouping said coefficients into a plurality of groups each containing coefficients having approximately the same significance and for selecting all coefficients from only selected ones of said groups.

16. The invention defined in claim 15, wherein said computing means includes

(a) means for computing estimates of motion of objects in said frames, and

(b) means for predicting the values of said pels based upon said motion estimates and the values of corresponding pels in at least one of said reference frames.

17. The invention defined in claim 15, wherein said computing means includes

(a) means for computing estimates of motion of objects in said frames, and

(b) means for interpolating the values of said pels based upon said motion estimates and the values of corresponding pels in at least two of said reference frames.

18. Apparatus (FIG. 2) for encoding a video signal representing the actual intensity values of picture elements (pels) in a series of frames, comprising

(i) means (210-228) for computing estimates of the intensity values of pels in at least two intermediate frames which lie between reference frames, based on the intensity values of pels in at least of said reference frames;

(ii) means (205) for combining error values for corresponding blocks of pels in each of said intermediate frames to form a series of transform coefficients representing said error values, each of said error values representing the difference between said estimated intensity value and said actual intensity value of said pel,

(iii) means (207) for selecting ones of said coefficients for further processing based upon selected characteristics of said coefficients, and

(iv) means (208) for encoding said selected coefficients to produce a representation of said video signal,

wherein said intensity estimate computing means includes

(a) means (228) for forming estimates of the motion of objects in said frames, and

(b) means (215 and

     217) for calculating said intensity estimates as a function of said motion estimates.

19. Apparatus for encoding a signal representing the intensity values of picture elements (pels) in a series of frames, comprising

means for computing error values s(m,n,p) for three dimensional blocks of pels, said 3D blocks including M×

N pels in each of P of said frames;

means for forming M×

N×

P discrete cosine transformation coefficients c(j,k,l) for each of said 3D blocks of error values; and

means for grouping said coefficients into S groups, each of said groups containing coefficients having approximately the same energy and selecting coefficients only from ones of said groups to represent said signal.

20. A method of encoding a video signal representing the acutal intensity values of picture elements (pels) in a series of frames, comprising the steps of

(i) computing estimates of the intensity values of pels in at least two intermediate frames which lie between reference frames, based on the intensity values of pels in said reference frames and on motion of objects in said frames;

(ii) combining error values for blocks of pels to form a series of transform coefficients representing said error values, each of said error values representing the difference between said estimated intensity value and said actual intensity value of said pel,

(iii) selecting ones of said coefficients for further processing based upon selected characteristics of said coefficients, and

(iv) encoding said selected coefficients to produce a representation of said video signal,

wherein said combining step includes;

(a) calculating DCT coefficients for two dimensional blocks of M×

N pels per some blocks of said frames; and

(b) calculating DCT coefficients for three dimensional blocks of M×

N pels in corresponding locations in each of P frames for other blocks of said frames.

21. The invention defined in claim 20 further including the step of adaptively switching between said last mentioned calculating steps.